LOGAN — While most people consider mosquitos an annoying pest, Utah State University Assistant Biology Professor Norah Saarman wanted to examine how they can spread infectious diseases.
“Our goal was to use images from space to see if we could predict how distant genetic mosquitos were or are across the landscape,” Saarman said. “What’s new about this is that we combined several approaches, one of those is called machine learning and it’s a really flexible way to ask if you can predict data.”
Saarman was also working to study the genetic connectivity of Aedes aegypti, an invasive species to North America that’s become widespread in the United States.
With Evlyn Pless of the University of California, Davis and Jeffrey Powell, Andalgisa Caccone and Giuseppe Amatulli of Yale University, Saarman published findings from a machine-learning approach to mapping landscape connectivity in the February 22, 2021 issue of the Proceedings of the National Academy of Sciences (PNAS). The team’s research was also supported by the National Institutes of Health.
“We’re excited about this approach, which uses a random forest algorithm that allows us to overcome some of the constraints of classical spatial models,” Saarman says. “Our approach combines the advantages of a machine-learning framework and an iterative optimization process that integrates genetic and environmental data.”
The most significant piece of the study, Saarman said, was the researcher’s method used: putting paper in a cup, waiting for mosquitos to lay eggs on top of the paper and then combining machine learning with a more interactive approach.
“The unique thing about this paper is that we used machine learning in a specific way for the first time,” Saarman said. “You can take the satellite data and try to predict the genetic data and then use the output from that to act like a feed for genetic iteration. It’s just a new way of combining two different methods.”
In its native Africa, Aedes aegypti was a forest dweller, drawing sustenance in landscapes uninhabited or scarcely populated by humans. The mosquito has since specialized to feed on humans, and thrives in human-impacted areas, favoring trash piles, littered highways and well-irrigated gardens.
“Using our machine-learning model and NASA-supplied satellite imagery, we can combine this spatial data with the genetic data we have already collected to drill down into very specific movement of these mosquitoes,” Saarman says. “For example, our data reveal their attraction to human transportation networks, indicating that activities such as plant nurseries are inadvertently transporting these insects to new areas.”
Public officials and land managers once relied on pesticides, including DDT, to keep the mosquitoes at bay.
“As we now know, those pesticides caused environmental harm, including harm to humans,” she said. “At the same time, mosquitos are evolving resistance to the pesticides that we have found to be safe for the environment. This creates a challenge that can only be solved by more information on where mosquitos live and how they get around.”
Saarman added the rugged survivors are not only adapting to different food sources and resisting pesticides, they’re also adapting to varied temperatures, which allows them to expand into colder ranges.
Current methods to curb disease-carrying mosquitoes focus on biotechnological solutions, including cutting-edge genetic modification.
“We hope the tools we’re developing can help managers identify effective methods of keeping mosquito populations small enough to avoid disease transmission,” Saarman said. “While native species play an important role in the food chain, invasive species, such as Aedes aegypti pose a significant public health risk that requires our vigilant attention.”